Artificial intelligence models, including those powering popular chatbots, can now predict an individual’s personality, behaviors, and daily emotional states with accuracy comparable to, and sometimes exceeding, that of their closest relationships, according to a new study published in the journal Nature Human Behavior.
The research, led by Aidan Wright, a professor of psychology and psychiatry at the University of Michigan, demonstrates the capacity of large language models (LLMs) to act as “judges” of personality by analyzing an individual’s own words. Researchers tasked AI programs like ChatGPT and Claude with assessing personality traits based on personal narratives – either short daily video diaries or longer recordings of spontaneous thoughts. The study encompassed data from over 160 participants gathered in both real-life and laboratory settings.
The AI-generated personality assessments closely mirrored self-assessments, and frequently aligned more accurately with individuals’ perceptions of themselves than evaluations provided by friends or family members. This performance significantly surpassed that of traditional text-analysis methods, according to the study.
“We were taken aback by just how strong these associations were, given how different these two data sources are,” Wright said.
Beyond simply mirroring self-perception, the AI’s personality ratings also proved predictive of real-world aspects of participants’ lives, including their emotional states, stress levels, social behaviors, and even diagnoses or treatment history related to mental health conditions. This suggests that personality is inherently expressed in everyday language, even when individuals are not consciously attempting to articulate their character.
Chandra Sripada, a professor of philosophy and psychiatry at the University of Michigan, noted that the findings reinforce the long-held belief that language contains significant clues about underlying psychological traits. He added that the availability of generative AI provides researchers with unprecedented tools to analyze open-ended writing and speech data with speed and precision.
The study acknowledges several limitations. The research relied on self-reported personality data and did not directly compare AI assessments to judgments from individuals’ social circles. The potential for variations in results across different demographic groups – such as age, gender, and race – remains unexplored. Researchers also have not yet determined whether AI and human assessments rely on the same underlying cues, or if AI could eventually surpass self-reports in predicting significant life outcomes, such as relationship success, educational attainment, health, or career progression.
Colin Vize, an assistant professor of psychology at the University of Pittsburgh, stated that the study “shows that AI can reliably uncover personality traits from everyday language, pointing to a new frontier in understanding human psychology.” Whitney Ringwald, an assistant professor of psychology at the University of Minnesota, emphasized that the results “really highlight how our personality is infused in everything we do, even down to our mundane, everyday experiences and passing thoughts.”
The rapid development of LLMs, including ChatGPT, Claude, Gemini, Llama, and Mistral, has transformed the AI landscape in recent years, according to a report from mybits.de published in November 2025. These tools offer potential benefits across various applications, from content creation to customer service automation and complex data analysis, but require careful selection based on specific leverage cases.
